Diagnostic and societal impact of implementing the syncope guidelines of the European Society of Cardiology (SYNERGY study)

We conducted a prospective multicentre pre-post study (Netherlands Trial Register, NTR6268 (old register)/https://​onderzoekmetmens​en.​nl/​nl/​trial/​23988 (new register)) comparing usual care (i.e. period prior to intervention), with the period after the syncope guidelines intervention. Five Dutch hospitals participated, including one university hospital (Leiden University Medical Centre, Leiden) and four regional ones (Diakonessenhuis, Utrecht; Rijnstate Hospital, Arnhem; Gelre Hospital, Apeldoorn; and Maasstad Hospital, Rotterdam).

The syncope guidelines intervention included two components: (1) implementation of the ESC 2018 Syncope Guidelines in the ED for all specialties by educating all involved healthcare providers and (2) establishment of a multidisciplinary syncope unit with quick referral routes for outpatient evaluation if needed [10].

Patients were enrolled between March 2017 and September 2019 and followed for 1 year. Patients were included 24 h per day and 7 days a week. The study started sequentially in each centre, with intervals of 2 to 4 months between each centre. The usual care and intervention periods lasted 6 months or longer to reach the inclusion target per centre. After completion of the usual care period, we paused the study enrolment for several weeks to educate the ED staff. All patients ≥ 18 years with suspected syncope visiting the ED were eligible for inclusion. We provided an operational definition for ‘suspected syncope’ to standardise clinical recognition in the ED. Suspected syncope was defined as TLOC that was (1) not due to traumatic head injury, (2) with no characteristics specific for seizure (e.g. post-ictal confusion, lateral tongue bite, > 20 jerks) and (3) with no characteristics specific for psychogenic TLOC (e.g. long duration, eyes closed during TLOC) [1, 29]. The exclusion criteria were (1) serious life-threatening conditions (e.g. massive bleeding; pulmonary embolus), (2) inability to give consent, (3) presyncope only and (4) syncope evaluation in the ED in the previous year. We also excluded cases in the usual care group who were referred to a syncope unit outside the study region as this included a component of the intervention. Participants were recruited 24 h per day and 7 days per week. We did not employ paid research staff. The PI of every site reminded all residents and specialists at daily rounds of the study. Residents and specialists recruited study participants. If a patient expressed interest, the patient received an information letter and a notification was sent to the researcher who contacted the subject after at least 24 h to provide additional explanation and to ask the subject to sign the informed consent letter if the subject agreed to participate.

We used 1-year follow-up data to determine the gold standard final diagnosis called the ‘reference diagnosis’ (Fig. 1) [32]. We first gathered all clinical information pertaining to the 1 year of follow-up. All participants received a questionnaire at baseline and at 3, 6 and 12 months of follow-up including questions on TLOC recurrence and doctor visits. If the patient reported TLOC recurrence evaluated in a non-participating hospital or by a general practitioner, we retrieved the clinical notes. A research nurse presented all medical information in a case record form stripped of identifying data to ensure blinding.

Next, cases were evaluated by applying the ESC criteria. All cases were evaluated by the research physician (MG) and, in case of doubt, by RDT. If the initial and any recurring events met the ESC criteria for a certain or highly likely diagnosis for reflex syncope or syncope due to orthostatic hypotension, those were used as reference diagnosis. When the treating physician issued a diagnosis of psychogenic TLOC, epilepsy or cardiac syncope, we reviewed the notes to assess whether definite proof was obtained: cases with psychogenic TLOC and epilepsy were evaluated by a neurologist considering video EEG as definite proof and cardiac syncope by a cardiologist considering ECG or echocardiographic findings as definite proof.

The next step involved the application of the ESC syncope risk stratification to cases without a definite diagnosis (Fig. 1). If patients had low-risk criteria only, without any recurrence during follow-up suggesting another diagnosis, the event was classified as ‘syncope due to reflex syncope or orthostatic hypotension’. Reflex syncope and orthostatic hypotension (OH) were bundled as one major ‘blood pressure related’ category, in view of common features in pathophysiology and risks.

Finally, if the data did not allow definite proof, the case was reviewed by the expert committee. Committees differed per occasion but always comprised one cardiologist (MB, MEWH or FdL), one neurologist (RDT, JGvD, IvR or EK) and one internist (JR or MPMH). All members had extensive experience in syncope care. All cases were first assessed independently by members of the expert committee, using predefined categories (Additional file 1): when they all agreed, their opinion was used as the final diagnosis; if not, consensus was reached in face-to-face or virtual meetings. We used Krippendorff’s alpha test to estimate the inter-rater agreement for the three raters in each consensus panel [33].

Our primary outcome measure was diagnostic accuracy. Diagnostic accuracy was defined as the proportion of cases with a correct diagnosis of TLOC within 1 year of follow-up. We determined the accuracy of the diagnosis by evaluating the concordance between the treating physician’s diagnosis (as described above) and the reference diagnosis. We only labelled diagnoses as discordant if the inconsistency occurred at the level of major TLOC categories (i.e. blood pressure-related causes of syncope, including reflex syncope and all forms of OH; cardiac syncope; epilepsy; psychogenic TLOC or TLOC of unknown cause; Additional file 1). For example, if the reference diagnosis was reflex syncope and the treating physician’s diagnosis was psychogenic TLOC, we labelled the diagnoses as discordant (i.e. inaccurate). Inaccuracies within major TLOC categories, such as orthostatic hypotension and reflex syncope (Additional file 1), were treated as accurate in our primary analysis of diagnostic accuracy. We also examined the accuracy of the subclassification of blood pressure-related syncope by the treating physician as a secondary outcome measure.

Secondary outcome measures included diagnostic yield, syncope-related admissions, tests, healthcare costs, societal costs and quality-adjusted life years (QALYs) in the year following the initial ED visit. We determined the diagnostic yield by calculating the proportion of patients who received a diagnosis of TLOC within 1 year of follow-up in each cohort (usual care and syncope guidelines intervention). To calculate the syncope-related healthcare costs and societal costs, we derived healthcare utilisation data from the patient records of the hospital where the initial ED visit took place. We also calculated healthcare costs in other hospitals, GP care and days lost from paid work to syncope with questionnaires, filled out by the patients at 3, 6 and 12 months. We selected all healthcare costs (e.g. diagnostic tests, therapies, hospital visits and admissions) directly related to the management of syncope (e.g. ECG) as well as indirect syncope-related healthcare costs (e.g. X-ray for trauma due to syncope). The latter category also included travel costs. Syncope-related societal costs were calculated by adding the costs of absenteeism to the healthcare expenditures. All costs were valued according to Dutch guidelines for economic healthcare evaluations, using Dutch reference prices if available or prices at the LUMC [34]. We reported all costs in Euros, at the price level of 2021. We calculated utilities from the EQ-5D-5L using the Dutch tariff, representing the value of quality of life on a scale from 0 (as bad as death) to 1 (perfect health) [35]. QALYs were calculated for each patient as the area under the utility curve.

The five EDs together saw ~ 166,000 patients per year (Leiden University Medical Centre: 30,000; Rijnstate Hospital: 38,000; Maasstad Hospital: 45,000; Gelre Hospital: 27,000; Diakonessenhuis: 26,000). A 1% syncope prevalence in the ED would result in ~ 1660 eligible patients per year [2‐5, 18]. We anticipated a diagnostic yield of 64% for usual care [32] and 80% for the intervention [10]. Assuming similar accuracy for both groups and an intervention effect of 16% (SD 3), we expected to detect significant differences across the five participating EDs with a power of 0.98 and on average across the general population with a power of 0.8 by including 275 patients per arm (two-sided alpha of 0.05).

We used IBM SPSS Statistics 23 for diagnostic analyses. Continuous data were expressed as mean (standard deviation) or median (interquartile range) when appropriate and categorical data as counts (proportion). We analysed the between-group differences in continuous data using Student’s t-test or non-parametric tests where appropriate, and we applied a logistic regression model to estimate the (log) risk ratio to compare diagnostic accuracy between both groups (syncope guidelines intervention and usual care) with fixed effects per centre. We used Stata/IC 14.2 for Windows to compare costs, using regression analysis with multiple imputation to account for missing questionnaire data (100 imputed data sets, with random effects per centre). We expressed the effect sizes for the main outcome measures as 95% confidence intervals. Statistical significance was set at p < 0.05.

A total of 548 patients gave informed consent between 1 March 2017 and 1 September 2019 (Additional file 2). Fourteen patients in the usual care group were excluded because they visited a syncope unit. Thirteen patients were excluded because in retrospect they had presyncope, leaving 275 patients in the usual care group and 246 patients in the syncope guidelines group. Most patients were 50 years or older (mean age 63 ± 17 in the usual care group and 64 ± 16 in the syncope algorithm group) and had a diagnosis of reflex syncope or OH according to the reference standard. The distribution of age, gender and reference diagnosis was comparable between the groups (Table 2).

Five non-syncope-related deaths were recorded during follow-up: three in the usual care group (due to choking, gastrointestinal bleeding and urosepsis later during the follow-up) and two in the syncope guidelines group (postoperative brain infarction and postoperative pulmonary embolus).

The diagnosis of the treating physician matched the reference diagnosis more frequently in the syncope guidelines group (n = 211, 86%) than in the usual care group (n = 191, 69%) (risk ratio 1.15 (95% CI 1.07 to 1.23; Fig. 2). The concordance between the treating physician’s diagnosis and the reference diagnosis improved across all major diagnostic categories with most marked improvements for those with blood pressure-related syncope (Figs. 3 and 4, Additional file 3). Diagnostic yield was higher (thus the proportion of unexplained syncope lower) in the syncope guidelines group compared to usual care (89% vs. 76%; 95% CI of the difference 6 to 19%).

The proportion of patients admitted directly from the ED was higher in the usual care group than in the syncope guidelines intervention group (53 vs. 43%; 95% CI of the difference: 2 to 19%). The reduction was most marked among those with exclusively low-risk criteria (usual care 43%; syncope guidelines 34%). Screening for OH in the ED was more frequently performed in the syncope guidelines group compared to usual care (26% vs. 16%, 95% CI of the difference: 3 to 17%). Fifteen per cent of all cases in the syncope guidelines group were referred to the syncope unit. We observed a modest increase in tilt table testing in the syncope guidelines group compared to usual care (7% vs. 2%; 95% CI of the difference: 2 to 9%). The proportion of cases referred for cardiac evaluations (Echocardiogram, Holter-EKG or implantable loop recorders) did not differ between the groups.

A total of 271 of the 521 cases met the ESC criteria for a certain or highly likely diagnosis for reflex syncope or syncope due to orthostatic hypotension. The definite diagnosis of the treating physician of cardiac syncope met our predefined criteria in all 13 cases. The remaining 247 cases without definite diagnosis were evaluated by our blinded consensus panel. The median interrater agreement of all five rounds of the consensus panel yielded a Krippendorf’s alfa statistic of 0.35 (range 0.27–0.41). In 102 out of 247 cases, all three independent raters unanimously agreed. In 135 cases, the final diagnosis was issued following a consensus meeting.

The questionnaire data were complete in 246 cases; in 255 cases, we lacked one or more follow-up questionnaires (108 one, 44 two, 108 three). Syncope-related healthcare costs did not significantly differ between the usual care group (€2393 ± 3844) and syncope guidelines group (€2035 ± 3353, p = 0.3) (Table 3), yet the total syncope-related societal costs were lower in the syncope guidelines group compared to usual care (saving of €908 per patient; 95% CI of the difference: €34 to €1782). The QALYs in the year following the initial ED visit were comparable between the groups.

Implementation of the ESC 2018 Syncope Guidelines in the ED with appropriate referral routes to a multidisciplinary syncope unit was associated with a higher diagnostic yield and accuracy and lower syncope-related societal costs.

A major strength of our work includes the combined diagnostic and economic evaluation of the ESC 2018 Syncope Guidelines with a 1-year follow-up in a real-world setting. Our intervention lacked supervised management to control the uptake of the ESC guidelines. As a result, we saw indications of the intervention in some but not all cases. For example, standing to upright blood pressure (BP) measurements in the syncope guidelines group were performed in only about a quarter of all cases, even though the number was significantly higher than in the usual care group. Similarly, the rate of hospital admission declined significantly in the guideline group, but the proportion of admissions remained higher than recommended. Previous studies using obligatory adherence to the guidelines demonstrated a higher increase of diagnostic yield and lower admission rates compared to our work [20, 25] underscoring that the added value of the ESC Syncope Guidelines implementation may be even higher. While the lack of supervised management may be considered a limitation, it reflects a real-world situation and strengthens the generalisability of our findings.

The intervention phase started well after the usual care phase, allowing possible bias due to changing circumstances. The alternative would require parallel pathways, but the two management strategies would then probably not remain separate. We compared diagnoses according to the reference standard in both cohorts and did not identify marked contrasts. Notably, the proportion of cases with cardiac syncope did not differ between both cohorts. We therefore assume that the sample bias was minimal.

We did not ask treating physicians to assign the likelihood of the diagnosis, as we expected that this would introduce an element of the intervention in the usual care period. Prospective evaluation of diagnostic likelihood may also introduce marked variation between physicians, particularly in the context of events with potential serious outcomes where even low risks may prompt ancillary testing to avoid adverse outcomes. Instead, we established criteria to evaluate the management plan and only considered syncope as explained if the final conclusion of the treating physician matched the treatment plan. Another limitation is the lack of a control group to control for secular trend bias [36]. We also considered the subjectivity of the diagnosis as a potential but inevitable limitation. This subjectivity is reflected by the relatively low interrater agreement. We therefore only accepted the expert opinion as a final diagnosis if all three independent raters unanimously agreed. In all other situations, we established multidisciplinary expert groups to obtain consensus. We only included cases with suspected syncope. Although near-syncope is known to confer similar risks for serious clinical events as syncope, we excluded cases with isolated presyncope as we anticipated difficulties in reliably identifying these cases in view of the nebulous nature of the presentation [37].

Healthcare use was assessed using a combination of electronic patient records and patient questionnaires, but we may still have missed some of the expenses in this diverse patient population. Our study is limited to the Dutch healthcare system. Although healthcare savings may differ between countries, the potential of ESC guideline-directed syncope care to save costs has been demonstrated in various healthcare systems [16, 20, 38]. A strength and novelty of our trial is that we assessed the impact of syncope on the absence from work, which emerged as an important driver in the lower syncope-related societal costs in the syncope guidelines group. Another strength is the long-term follow-up incorporating the post-ED stratification workup, allowing us to evaluate the precision of the suggested diagnosis. Our intervention included two components that were implemented simultaneously, so we could not determine the impact of the syncope guidelines implementation and syncope unit establishment separately. We saw, however, signs of both interventions as evaluations that are strongly recommended by the syncope guidelines (e.g. OH measurements) significantly increased in the implementation phase as did the number of syncope unit referrals. Another sign of our intervention concerned the improved diagnostic precision. Our data indicate that the proportion of cases in need for syncope unit referral was ~ 15% [31] confirming the notion that the initial evaluation is sufficient in the majority of cases [1‐6, 32, 39]. We need, however, to keep in mind that our trial presents the results immediately following a service delivery intervention. We expect that the referral rate and hereby the added value of the syncope unit on the emergency evaluation of syncope may increase over time.

Our findings go beyond previous implementation studies as no studies investigated the simultaneous implementation of the syncope guidelines and quick referrals to a syncope unit [6, 22, 40]. A multicentre study in Italy showed that a structured diagnostic algorithm at the ED based on the syncope guidelines improved the diagnostic yield significantly [19], but without a control group. Another Italian study showed that a standardised care pathway led to a higher diagnostic yield than usual care [20]. Although the study included a control group, the intervention only comprehended the ED evaluation. A recent Dutch study showed that evaluation in the ED by a research physician with strict adherence to the ESC guidelines improved the proportion of correct diagnoses from 65 to 80% [25]. An educational programme promoting the ESC guidelines had only a limited impact on syncope care and did not improve cost-effectiveness, thus underscoring the complexity of guideline implementation in the ED [41]. While education also constituted a key element of our intervention, we additionally offered quick referral routes for patients with intermediate risk. This might explain why our study yielded better results than the former educational intervention study. In a single-centre study in Ireland, ESC Syncope Guidelines implementation was carried out by introducing a local Integrated Care Plan [23]. Local physicians were instructed several times prior to implementation and during 4 weeks afterwards. This resulted in increased appropriate referral rates to the syncope unit. The FAST II study demonstrated the added value of a tertiary syncope unit by demonstrating a high diagnostic yield, accuracy and safety [42, 43]. Notably, those who were diagnosed with reflex syncope underwent a median of nine tests prior to referral. The study also suggested a knowledge gap as often no ancillary testing was needed to yield a highly reliable diagnosis despite multiple consultations by various specialties. Initial orthostatic hypotension was identified as a frequent blind spot which is in line with the improved diagnostic yield for BP-related causes of syncope we found following ESC Syncope Guidelines implementation [42‐44]. Similarly, a dedicated outpatient evaluation including tilt testing in cases with unexplained syncope presenting to the ED in Malmö considerably improved diagnostic efficacy and accuracy [45]. As in our study, the majority of cases in this cohort were found to have BP-related causes of syncope including vasovagal syncope, carotid sinus hypersensitivity and orthostatic hypotension. We studied the BP-related causes as a composite outcome as we anticipated difficulties in classification due to incomplete evaluation (e.g. no standing test) or incomplete records (e.g. no mention of the presence or absence of autonomic activation). Although our primary analysis focussed on the accuracy of identifying the major TLOC groups, the classification of BP-related syncope matched the reference diagnosis more frequently in the syncope guidelines group compared to usual care (Additional file 1). In line with previous studies, we found that hospitalisation was the primary driver of syncope-related healthcare costs [2‐5, 39]. Annual admission costs exceeded $2.4 billion in the USA, as high as those for HIV and asthma [46, 47]. A recent survey reported higher admission rates among patients with lower-risk syncope in regions with higher malpractice claims, suggesting defensive medicine [4]. While the admission rate significantly declined following our intervention, the overall syncope-related health costs between the groups did not reach significance, probably due to a slight increase in outpatient evaluations in the syncope guidelines group [47]. As our study was powered for diagnostic outcomes, the lack of statistical significance of differences between types of healthcare costs and absenteeism does not allow strong conclusions. However, these differences were mostly in favour of our intervention, and the combined syncope-related societal costs were significantly lower following our intervention.

Previous work suggested that guideline adherence is promoted by combining knowledge transmission, reflective practice and a supportive environment [39, 48]. We, as did others, [49, 50] found implementation of the ESC guidelines to be complex, requiring tailored strategies to overcome potential barriers. While most professionals in our study welcomed the structured workflow in the ED and the multidisciplinary syncope unit as a useful solution to a perceived need in clinical practice, remaining barriers occurred on the level of the individual healthcare professional (e.g. inexperienced residents having to work with the guideline in the ED) and the organisational context (e.g. specialists not relinquishing preceding procedures) [51]. Time pressure emerged in our study as one of the biggest obstacles at the organisational level, while orthostatic BP measurements were often not performed because their yield was perceived as low [51]. The low rate of orthostatic BP measurements in the usual care group (16%) is in line with a previous Dutch survey (16%) [52]. In the latter survey, orthostatic BP measurements were only performed in 48% of those who received a diagnosis of OH, thus questioning the validity of this diagnosis [52]. A US survey found that implementation of new syncope guidelines was seen as a burden, potentially decreasing compliance [49]. Change may be more likely if implementation strategies are specifically chosen to address frequently encountered barriers in physicians’ existing knowledge, attitudes and behaviour. Reliable risk prediction tools are also much needed as they could help to support clinical judgement. Many syncope risk stratification scores have been proposed, yet with conflicting results [6]. A recent multicentre trial reported that early standardised clinical judgement for cardiac syncope in the ED outperformed the frequently used Evaluation of Guidelines in Syncope Study (EGSYS) syncope prediction score [53]. History taking is still the cornerstone of the initial syncope evaluation [6, 31, 54].